From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001 From: Jianfeng Liu Date: Thu, 15 May 2025 17:32:53 +0800 Subject: blink: fix lsx code of webgl_image_conversion_lsx --- .../loongarch64/webgl_image_conversion_lsx.h | 88 +++++++++---------- 1 file changed, 44 insertions(+), 44 deletions(-) diff --git a/third_party/blink/renderer/platform/graphics/cpu/loongarch64/webgl_image_conversion_lsx.h b/third_party/blink/renderer/platform/graphics/cpu/loongarch64/webgl_image_conversion_lsx.h index 114f3c112e..1e2096f651 100644 --- a/third_party/blink/renderer/platform/graphics/cpu/loongarch64/webgl_image_conversion_lsx.h +++ b/third_party/blink/renderer/platform/graphics/cpu/loongarch64/webgl_image_conversion_lsx.h @@ -91,19 +91,19 @@ ALWAYS_INLINE void PackOneRowOfRGBA8LittleToRA8(const uint8_t*& source, v4f32 mask_falpha = __lsx_vffint_s_w(mask_lalpha); v16u8 ra_index = {0,19, 4,23, 8,27, 12,31}; for (unsigned i = 0; i < pixels_per_row_trunc; i += 4) { - v16u8 bgra = *((__m128i*)(source)); + __m128i bgra = *((__m128i*)(source)); //if A !=0, A=0; else A=0xFF - v4f32 alpha_factor = __lsx_vseq_b(bgra, mask_zero); + __m128i alpha_factor = __lsx_vseq_b(bgra, mask_zero); //if A!=0, A=A; else A=0xFF alpha_factor = __lsx_vor_v(bgra, alpha_factor); alpha_factor = __lsx_vsrli_w(alpha_factor, 24); - alpha_factor = __lsx_vffint_s_w(alpha_factor); - alpha_factor = __lsx_vfdiv_s(mask_falpha, alpha_factor); + alpha_factor = (__m128i) __lsx_vffint_s_w(alpha_factor); + alpha_factor = (__m128i) __lsx_vfdiv_s((__m128) mask_falpha, (__m128) alpha_factor); - v16u8 component_r = __lsx_vand_v(bgra, mask_lalpha); - component_r = __lsx_vffint_s_w(component_r); - component_r = __lsx_vfmul_s(component_r, alpha_factor); - component_r = __lsx_vftintrz_w_s(component_r); + __m128i component_r = __lsx_vand_v(bgra, mask_lalpha); + component_r = (__m128i) __lsx_vffint_s_w(component_r); + component_r = (__m128i) __lsx_vfmul_s((__m128) component_r, (__m128) alpha_factor); + component_r = __lsx_vftintrz_w_s((__m128) component_r); v2u64 ra = __lsx_vshuf_b(bgra, component_r, ra_index); __lsx_vstelm_d(ra, destination, 0, 0); @@ -138,19 +138,19 @@ ALWAYS_INLINE void PackOneRowOfRGBA8LittleToR8(const uint8_t*& source, v4u32 mask_lalpha = __lsx_vreplgr2vr_w(0x0ff); v4f32 mask_falpha = __lsx_vffint_s_w(mask_lalpha); for (unsigned i = 0; i < pixels_per_row_trunc; i += 4) { - v16u8 bgra = *((__m128i*)(source)); + __m128i bgra = *((__m128i*)(source)); //if A !=0, A=0; else A=0xFF - v4f32 alpha_factor = __lsx_vseq_b(bgra, mask_zero); + __m128i alpha_factor = __lsx_vseq_b(bgra, mask_zero); //if A!=0, A=A; else A=0xFF alpha_factor = __lsx_vor_v(bgra, alpha_factor); alpha_factor = __lsx_vsrli_w(alpha_factor, 24); - alpha_factor = __lsx_vffint_s_w(alpha_factor); - alpha_factor = __lsx_vfdiv_s(mask_falpha, alpha_factor); + alpha_factor = (__m128i) __lsx_vffint_s_w(alpha_factor); + alpha_factor = (__m128i) __lsx_vfdiv_s((__m128) mask_falpha, (__m128) alpha_factor); - v16u8 component_r = __lsx_vand_v(bgra, mask_lalpha); - component_r = __lsx_vffint_s_w(component_r); - component_r = __lsx_vfmul_s(component_r, alpha_factor); - component_r = __lsx_vftintrz_w_s(component_r); + __m128i component_r = __lsx_vand_v(bgra, mask_lalpha); + component_r = (__m128i) __lsx_vffint_s_w(component_r); + component_r = (__m128i) __lsx_vfmul_s((__m128) component_r, (__m128) alpha_factor); + component_r = __lsx_vftintrz_w_s((__m128) component_r); component_r = __lsx_vpickev_b(component_r, component_r); component_r = __lsx_vpickev_b(component_r, component_r); @@ -172,41 +172,41 @@ ALWAYS_INLINE void PackOneRowOfRGBA8LittleToRGBA8(const uint8_t*& source, v4f32 mask_falpha = __lsx_vffint_s_w(mask_lalpha); v16u8 rgba_index = {0,1,2,19, 4,5,6,23, 8,9,10,27, 12,13,14,31}; for (unsigned i = 0; i < pixels_per_row_trunc; i += 4) { - v16u8 bgra = *((__m128i*)(source)); + __m128i bgra = *((__m128i*)(source)); //if A !=0, A=0; else A=0xFF - v4f32 alpha_factor = __lsx_vseq_b(bgra, mask_zero); + __m128i alpha_factor = __lsx_vseq_b(bgra, mask_zero); //if A!=0, A=A; else A=0xFF alpha_factor = __lsx_vor_v(bgra, alpha_factor); alpha_factor = __lsx_vsrli_w(alpha_factor, 24); - alpha_factor = __lsx_vffint_s_w(alpha_factor); - alpha_factor = __lsx_vfdiv_s(mask_falpha, alpha_factor); + alpha_factor = (__m128i) __lsx_vffint_s_w(alpha_factor); + alpha_factor = (__m128i) __lsx_vfdiv_s((__m128) mask_falpha, (__m128) alpha_factor); v16u8 bgra_01 = __lsx_vilvl_b(mask_zero, bgra); v16u8 bgra_23 = __lsx_vilvh_b(mask_zero, bgra); - v16u8 bgra_0 = __lsx_vilvl_b(mask_zero, bgra_01); - v16u8 bgra_1 = __lsx_vilvh_b(mask_zero, bgra_01); - v16u8 bgra_2 = __lsx_vilvl_b(mask_zero, bgra_23); - v16u8 bgra_3 = __lsx_vilvh_b(mask_zero, bgra_23); - - bgra_0 = __lsx_vffint_s_w(bgra_0); - bgra_1 = __lsx_vffint_s_w(bgra_1); - bgra_2 = __lsx_vffint_s_w(bgra_2); - bgra_3 = __lsx_vffint_s_w(bgra_3); - - v4f32 alpha_factor_0 = __lsx_vreplvei_w(alpha_factor, 0); - v4f32 alpha_factor_1 = __lsx_vreplvei_w(alpha_factor, 1); - v4f32 alpha_factor_2 = __lsx_vreplvei_w(alpha_factor, 2); - v4f32 alpha_factor_3 = __lsx_vreplvei_w(alpha_factor, 3); - - bgra_0 = __lsx_vfmul_s(alpha_factor_0, bgra_0); - bgra_1 = __lsx_vfmul_s(alpha_factor_1, bgra_1); - bgra_2 = __lsx_vfmul_s(alpha_factor_2, bgra_2); - bgra_3 = __lsx_vfmul_s(alpha_factor_3, bgra_3); - - bgra_0 = __lsx_vftintrz_w_s(bgra_0); - bgra_1 = __lsx_vftintrz_w_s(bgra_1); - bgra_2 = __lsx_vftintrz_w_s(bgra_2); - bgra_3 = __lsx_vftintrz_w_s(bgra_3); + __m128i bgra_0 = __lsx_vilvl_b(mask_zero, bgra_01); + __m128i bgra_1 = __lsx_vilvh_b(mask_zero, bgra_01); + __m128i bgra_2 = __lsx_vilvl_b(mask_zero, bgra_23); + __m128i bgra_3 = __lsx_vilvh_b(mask_zero, bgra_23); + + bgra_0 = (__m128i) __lsx_vffint_s_w(bgra_0); + bgra_1 = (__m128i) __lsx_vffint_s_w(bgra_1); + bgra_2 = (__m128i) __lsx_vffint_s_w(bgra_2); + bgra_3 = (__m128i) __lsx_vffint_s_w(bgra_3); + + __m128i alpha_factor_0 = __lsx_vreplvei_w(alpha_factor, 0); + __m128i alpha_factor_1 = __lsx_vreplvei_w(alpha_factor, 1); + __m128i alpha_factor_2 = __lsx_vreplvei_w(alpha_factor, 2); + __m128i alpha_factor_3 = __lsx_vreplvei_w(alpha_factor, 3); + + bgra_0 = (__m128i) __lsx_vfmul_s((__m128) alpha_factor_0, (__m128) bgra_0); + bgra_1 = (__m128i) __lsx_vfmul_s((__m128) alpha_factor_1, (__m128) bgra_1); + bgra_2 = (__m128i) __lsx_vfmul_s((__m128) alpha_factor_2, (__m128) bgra_2); + bgra_3 = (__m128i) __lsx_vfmul_s((__m128) alpha_factor_3, (__m128) bgra_3); + + bgra_0 = __lsx_vftintrz_w_s((__m128) bgra_0); + bgra_1 = __lsx_vftintrz_w_s((__m128) bgra_1); + bgra_2 = __lsx_vftintrz_w_s((__m128) bgra_2); + bgra_3 = __lsx_vftintrz_w_s((__m128) bgra_3); bgra_01 = __lsx_vpickev_b(bgra_1, bgra_0); bgra_23 = __lsx_vpickev_b(bgra_3, bgra_2); -- 2.51.0